ChatGPT's New Image Generator Exceeds Expectations, Now Free to Try

If you're a fan of AI-generated images, you've probably heard of OpenAI's DALL-E model. However, since its initial release, other models have surpassed it in capabilities. So, when OpenAI launched their latest model, GPT-4o, I was initially skeptical. But after putting it to the test, I must say, I've been pleasantly surprised.

Getting Started

When DALL-E first came out, it had its own dedicated website. But now, it's integrated into ChatGPT, which has brought a lot of convenience to the table. You can now request images in the same interface where you're already chatting, making the process seamless and eliminating the need to switch contexts.

With the introduction of GPT-4o, OpenAI has maintained this user-friendly approach. The default image generator for paid subscribers has been switched from DALL-E to GPT-4o. This made it incredibly easy for me to start creating images using my ChatGPT Plus account. All I had to do was type in my prompt, and voila! The images were generated. You can also access it via the Sora interface, if you prefer.

Even free users can generate images, though there was a brief hiccup at launch. Initially, it was announced that the model would be available to all users, but then CEO Sam Altman announced a delay for the free tier. Fortunately, it became available to free users just a week later. However, there's a catch for free users: you need to use the shortcut "/create image" to activate GPT-4o. Simply typing a request like "Create an image of XYZ" will default to the older DALL-E model, which produces lower-quality images. I hit my daily limit after generating just three images from my free account, so if you want more access, ChatGPT Plus is the way to go.

The Images

Now, let's talk about the images themselves. Once you enter your prompt, the AI generates the image in under a minute. It takes a bit longer than before, but trust me, it's worth the wait. The images are packed with detail, texture, realism, and even accurate text. Instead of just describing it, I'll show you some examples.

Prompt: Can you generate a realistic image of a chameleon, up close, shot as if it were in National Geographic in 16:9 ratio?

Prompt: Can you generate an image of a laptop open on a desk that says, "This model is so good that it can even get text and hands right, which are usually major challenges for AI models," with hands typing on a keyboard in 16:9 ratio?

Prompt: Can you generate a realistic photo of a close-up of a woman in a crowd in Times Square looking at the camera and smiling, with the quality of one taken on a DSLR?

As you can see, the image generator excels at following the prompt and delivering high-quality, realistic images. But to truly gauge its performance, I compared it to other leading AI image generators like Midjourney, Google's Imagen 3, and Adobe Firefly using the same prompt.

Prompt: Can you generate an image of a vibrant, realistic hummingbird perched on a tree?

GPT-4o's rendition holds its own against these competitors, and it clearly outperforms the older DALL-E model.

Other Notable Features

While the quality of the images is a standout feature, there are other aspects that make GPT-4o a strong contender. One of the best parts is its integration within the chatbot's interface. This makes it easy to refine your image requests with simple, natural language prompts. Plus, since the chatbot retains the context of your conversation, it can incorporate that into your image.

For instance, if you're discussing a birthday party, you can simply ask, "Can you now create an invite that has the information above on it?" without having to repeat everything. I tried this while planning a housewarming party, and it worked seamlessly.

You can also upload reference images and ask ChatGPT to create variations or incorporate elements into new images. For example, you could upload a selfie and have it transformed into an anime style, much like Sam Altman did in his recent X post.

changed my pfp but maybe someone will make me a better one

— Sam Altman (@sama) March 26, 2025

These customization options are fantastic for creatives, who can also request transparent backgrounds or incorporate brand style guides like hex codes or logos.

Speaking of Altman, I was able to generate an image of him wearing a party hat, thanks to the model's more relaxed safeguards. The announcement blog post noted that these safeguards are especially robust around nudity and graphic violence when real people are involved.

I'm not sure if there's a practical use for this feature, but it's an interesting change worth noting. When I tried to create an image of Mickey Mouse, it was blocked due to copyright concerns, so not all public figures are fair game.

Overall

In conclusion, the GPT-4o image generator is a significant improvement over the DALL-E models and ranks among the best I've tested. Is it worth the $20 monthly subscription? If you're solely interested in high-quality image generation, there are free alternatives like Adobe Firefly or Google's Imagen 3 that are quite capable.

However, the updated features are now rolling out to all users, including those on the free tier. Just remember, free users need to use the "/create image" shortcut to access GPT-4o; otherwise, you'll get the lower-quality DALL-E results.

For frequent ChatGPT users, upgrading to ChatGPT Plus becomes more appealing. You'll get access to all of OpenAI's latest features, including high-quality image and video generation, for just $20 a month. This is a pretty good deal, especially when compared to other options like Midjourney, which starts at $10 per month and only offers image generation.

Want more stories about AI? Sign up for Innovation, our weekly newsletter.